交通安全是公共安全的重要组成部分，人因事故防范则是交通安全研究的重中之重。监测评估驾驶员状态并及时进行风险预警能够减少道路交通事故，提升驾驶安全。然而，现有研究在认知机制层面对驾驶员行为的解析尚显不足，驾驶员驾龄、驾驶风格等个体特征以及情绪等驾驶状态对风险的影响与机理不够明晰，驾驶员风险性的量化参数表征和综合安全评价方法也有待进一步探索。因此，本研究重点针对驾驶员行为及状态的风险特性探究与安全评价等科学问题，基于自然驾驶和模拟驾驶实验数据，对驾驶员的基本刺激响应行为特征进行了解析，对驾驶员的不同状态风险特性开展了探究；综合以上行为及状态风险特性分析结果，构建了风险行为驾驶模型，提出了驾驶员安全表征及评价方法。在驾驶员行为特征研究方面，基于道路自然驾驶实验行车数据与驾驶员近红外脑功能成像数据，构建了外界刺激与驾驶员大脑响应时序耦合模型，明确了驾驶员对距离、速度、加速度和车头时距等直观刺激量具有显著感知；从认知层面解析了驾驶员对道路环境刺激进行响应的脑区分布与时间延迟特性，为后续驾驶员行为模型的构建提供参考依据；研究了驾龄、职业、驾驶风格等个体特征对驾驶员感知与响应的影响，为最终的驾驶员安全性评价提供理论基础。在驾驶员状态风险机理研究方面，基于事故数据关键特征构建了风险驾驶场景，设计开展了不同状态模拟驾驶实验；通过小波变换定量研究了驾驶员状态及操作行为时频特征与车辆运动特性的关系，构建了驾驶状态作用于驾驶行为，驾驶行为控制车辆运动，车辆运动特性影响驾驶风险的耦合作用链；阐明了驾驶状态对风险的影响机理，为驾驶员安全评价奠定了数据基础，提供了实验支撑。在驾驶员安全评价研究方面，综合考虑驾驶员的风险感知和驾驶预期，构建了风险行为驾驶模型，并在实际道路行车数据集上得到验证；对驾驶员安全性进行了参数化表征，量化研究了驾驶行为特性对驾驶风险的影响；提出了基于行为特征的综合安全评价及风险识别方法，为驾驶风险的监测预警提供技术支撑。论文通过实验探究、理论建模、模拟仿真，对人工驾驶行为及状态的风险特性开展了解耦分析，进而构建了风险行为驾驶模型，进行了驾驶员安全性参数表征，并定量化探索了其对驾驶风险的影响规律及作用机理，为驾驶员行为及状态的风险特性研究与安全评估预警提供了理论、实验和方法支撑。

Traffic safety has consistently been a critical aspect of public safety. The prevention of accidents attributed to human factors has long been a vital subject in the field of traffic safety research. Driver state evaluation and timely risk warnings can reduce traffic accidents and enhance driver safety. However, existing research still needs more analysis of the cognitive mechanism of driving behavior. Deeper investigation is needed to study the influence and underlying mechanism of driver individual characteristics, such as driving experience and driving style, as well as driving states like emotional driving, on driving risk. Moreover, there is a necessity for further exploration in the quantification of parameters representing driver risk and the development of comprehensive driving safety evaluation method. Therefore, this study focuses on the scientific research of the risk characteristics and safety evaluation of driver behavior and state. Based on data from naturalistic driving experiment and simulated driving experiment, the basic driver stimuli-response behavioral pattern is analyzed, the risk characteristics under different driving state is investigated. Through comprehensive analysis, a risk behavior driving model is constructed and a method for driver safety characterization and evaluation is proposed.In the study of driver behavioral patterns, this paper constructs a temporal responses model of drivers to road stimuli based on naturalistic driving data and brain-image functional near-infrared spectroscopy (fNIRS) data of drivers. It is clarified that drivers have a significant perception and response to intuitive stimuli such as distance, velocity, acceleration and headway. The distribution of the brain regions and the time delay characteristics of the driver‘s response to road stimuli are deeply investigated from the cogonitive perspective, which can support the subsequent modeling of driver behavior. The influence of driver attributes, such as driving experience, profession, and driving style on driving perception and reaction, is also investigated, providing a theoretical basis for safety evaluation of drivers.In the study of the relationship between driving risk and driver states, risk-driving scenarios are built based on key feature of accident. Then, a simualted driving experiment under various driving states is designed and conducted. Through wavelet transform analysis, the relationship between driver states, behavioral patterns and vehicle movement characteristics in both time and frequency domains is quantitatively investigated. This analysis established a causal chain that driving states influence behaviors, behaviors control vehicle movement, and motion characteristics impact driving risks. The study explores the influence and mechanism of driver states on driving risk, which can serve as a foundational database and provide experimental support for the safety evaluation of driver states.In the study of driver safety evaluation, a risk behavior driving model is developed by taking into account both the risk perception and driving expectation characteristics of drivers. The model is calibrated and validated through simulation using real-world driving data. With the characteriazation and parameterization of driver safety, the influence of driving behavior patterns on driving risk is quantitatively investigated. Subsequently, a safety evaluation and risk detection method based on behavioral features is proposed, providing technological support for the monitoring and warning of driving risk.In summary, this paper systematically explore the risk characteristics inherent in human driving behavior and driving states through experimental analysis, theoretical modelling and simulation studies. The driving model of risk behavior is constructed, the safety characteristics of drivers is parameterized, and the mechanism influencing driving risk is quantitatively investigated. These work can provide theoretical, methodological and technological support for driver behavior modelling, driving risk detection and driver safety evaluation.